Abstract
Macroeconomics has been constrained by assumptions that the economy tends toward full employment, and longterm growth depends on exogenous factors, productivity, and labor force growth. This discounts effects of demand on the labor force participation and immigration, and on investment and productivity growth. Expansionary fiscal policy and greater effective demand could induce reverse (or benevolent) hysteresis where capacity grows faster with higher employment, investment, and productivity.
In spring 2016, I published a forecast of the economic impact of the economic program proposed by then-candidate Bernie Sanders, a program of redistribution and fiscal expansion on a scale without precedent in the United States since the New Deal and World War II. My estimates were immediately dismissed as promising “puppies and rainbows,” even “magic flying puppies with winning Lotto tickets tied to their collars,” and supported by “no credible economic research” (Calmes 2016; Krueger et al. 2016). As the debate progressed and more thoughtful critics articulated their objections to my model, it became clear that divergent views on two fundamental questions in political economy shaped the reaction to my analysis. First, do capitalist economies have a self-regulating tendency toward full employment? Second, is their growth capacity set exogenously without regard to aggregate demand or does economic capacity tomorrow depend on output today (Friedman 2016; Romer and Romer 2016)?
To clarify, we should consider different meanings of the word “capacity.” If we assume full employment, then an economy’s capacity output is the actual level of output attained. Alternatively, if we do not assume full employment, then capacity can be understood as the maximum output attainable if resources were fully utilized where we estimate the volume of these resources, the labor force, means of production, and technology, assuming either that they had all progressed at a rate set exogenously, or that they had progressed at a rate influenced by demand conditions. To repeat, assuming full employment joins these different meanings because actual output is potential output; if we assume full employment, then capacity grows at a rate set independently of demand because there are no fluctuations in demand. If we recognize fluctuations in demand, then we could still assume that capacity is set exogenously with respect to demand; but there is space for assuming endogenous capacity.
So demand may influence the growth of capacity if we reject both the full-employment assumption and the assumption of exogenous growth in labor force and productivity. A growing body of research supports this approach, including endogenous growth theories. Advocates of hysteresis have also found a connection between capacity growth and demand, at least on the downside (Lindbeck and Snower 1989; hysteresis should be positive as well as negative, point noted by Bernstein [2014] and elsewhere).
A gap has thus developed between advances in macroeconomic theory and applied work. Although many theorists accept that demand can influence capacity, that demand can create supply, many applied economists continue to assume a tendency toward full employment and treat changes in capacity as independent of demand. In their empirical analysis, they equate measured output with potential, and assume that slow growth reflects diminished capacity rather than underutilization. This has been the practice, for example, of the Congressional Budget Office (CBO) that measures capacity by actual output and, since the onset of the Great Recession (or the Lesser Depression), has repeatedly lowered estimates of capacity in line with low levels of employment and output (Summers presents a graphic showing these revisions in Teulings and Baldwin 2014: 28–29). In effect, they have assumed that output growth is slow because of exogenous conditions, and then argued that there is full employment because output is at that ex post defined low level of capacity.
Instead of inferring capacity from output by assuming that output is close to capacity, I propose a Keynesian alternative where output may diverge substantively from capacity for indefinite periods, and where capacity change is driven by the level of output. Thus, we move from the study of fluctuations, the original Keynesian question of whether the economy has a natural tendency toward full employment, to the larger question of levels and of economic growth: is capacity output independent of the level of demand, or can demand induce capacity? And can increasing effective demand produce a higher rate of growth in attainable capacity?
My Sanders project assumed the latter Keynesian position, one that not only has support from mainstream macroeconomists but draws on the work of some of my critics (DeLong 2014; Krugman 2013). I had, therefore, assumed that my contribution would be welcomed. It was a contribution to the argument that, by reversing the negative effects of hysteresis on capacity output, Keynesian pump priming can have lasting effects on output and on growth (see Ball 2014; Blanchard and Summers 1986; Fatás and Summers 2016; Stockhammer and Sturn 2012).
If business recessions “cast long shadows” by lowering investment, reducing labor-force attachment, and slowing technological progress, then an aggressive fiscal policy that produces a high-pressure full-employment economy should be able to promote higher growth (Fatás 2000). (Otherwise growth rates would long ago have fallen to zero!) Others have made similar arguments. Janet Yellen recently asked, “[a]re there circumstances in which changes in aggregate demand can have an appreciable, persistent effect on aggregate supply?” (Yellen 2016). Paul Krugman similarly suggests that “demand creates its own supply” (Krugman 2015). And while hedging that “the data are not conclusive,” Brad DeLong and Summers argue that by mitigating hysteresis effects and avoiding the “protracted output losses like those suffered by the United States in recent years,” an aggressive fiscal policy can “raise potential future output” (DeLong and Summers 2012: 237).
In the space here, I only sketch some of the pathways between demand and capacity and do not attempt a full empirical assessment. The purpose here is to open a discussion, to move us beyond talk of puppies and unicorns to a serious discussion of the sources of investment, productivity, and labor-force participation.
1. Components of Capacity: Investment
Beyond his concerns for stabilization policy, for limiting macroeconomic fluctuations and unemployment, John Maynard Keynes had larger concerns beyond achieving full employment for promoting investment and growth. Although he wrote the general theory during the Great Depression of the 1930s, his ideas grew out of his experience with slow growth and economic stagnation in Britain in the 1920s and earlier. Keynes did not intend for the general theory as only a study in stabilization policy but as a treatise on long-term economic policy leading toward the “euthanasia of the rentier, and, consequently, the euthanasia of the cumulative oppressive power of the capitalist to exploit the scarcity value of capital.” Effective monetary and fiscal policy, including public investment, was to maintain rates of employment and economic growth such that there will be incentives for sufficient investment and technological progress to drive down the rate of interest (Keynes 1926, 1936, chapter 24). Eventually, by promoting higher levels of investment and growth, activist policy will produce enough productivity growth to usher in an economic utopia where humanity would have solved the economic problem, the need to work (Keynes 1963).
For Keynes, countercyclical policy does more than raise employment; it promotes long-run growth, by encouraging income-sensitive investment in plant, equipment, and research and development (R&D). Working through an “accelerator,” higher levels of demand lead to investment that produce the higher capacity and output levels to encourage further output-sensitive investment. It is long established that investment in new plant and equipment is highly procyclical. Businesses invest in response to pressures on utilization; they substitute capital for relatively scarce labor when employment and wages rise; higher demand is often taken as a measure of future demand, and higher retained earnings during an economic expansion lower the effective cost of investment by reducing the need for outside funding. (Procyclical investment behavior is discussed by Chirinko [1993]. The cyclical effect of R&D is emphasized by Anzoategui et al. [2016]).
To argue for exogenous capacity growth and that aggressive demand policy has no effect on capacity, one must assume that investment is insensitive to demand, that it adds nothing to capacity, or that aggressive demand policy is ineffective. Few will accept any, not to mention all, of these assumptions. Instead, it is reasonable to assume that investments are productive, higher investment raises labor productivity, and, because investment is procyclical, it makes labor productivity procyclical as well. By creating a pattern of persistently low investment, slow economic growth contributes to a general slowdown in capacity and in economic growth; by raising demand, effective fiscal stimulus will encourage more investment and greater capacity growth.
Beyond illustrating the procyclical nature of investment, the data in Figure 1 indicate another, disturbing trend: a decline in investment over recent business cycle peaks. Declining investment in plant, equipment, and research means slower productivity growth. Similarly, an aggressive Keynesian policy could promote income and productivity growth by substituting government investment in infrastructure, education, and research for faltering private investment (Cohen and DeLong 2016).
Investment as share of GDP, quarterly data, 1980 to 2016.
2. Components of Capacity: Population and Labor-Force Participation
The question of demand-led growth also turns on the cyclical sensitivity of population and labor-force participation. Those who argue for exogenous capacity insist that demographics determine the labor force and estimate the size of the potential labor force by multiplying fixed labor-force participation rates to a demographically determined population age structure. Both are independent of demand.
2.1. Labor-Force Participation
The CBO projects that the labor force will increase by about 0.8 percent a year over the next decade, slower than the 1.2 percent growth in the 1990s or the 2.1 percent growth for the previous decades. This slowdown is in part due to slower population growth, and, for the rest, comes because the CBO assumes the decline in labor-force participation that began with the recession of 2000 and accelerated with the Lesser Depression of 2007 will continue (the “lesser depression” phrase comes from DeLong [2014]). The CBO assumes that the adult labor-force participation rate will fall from 64.7 percent in 2011 to 63.0 percent in 2021 because of population aging, the leveling off in participation among women, and a continued decline in participation among people below the age of 25 (Congressional Budget Office 2011: 11, 17).
To be fair, the CBO acknowledges that the labor force reflects macroeconomic circumstances and effective demand. In 2011, it noted that labor-force participation rates were low by historic standards and projected an increase with economic recovery.
I would argue that this seriously understates the capacity for labor-force growth. As of September 2016, the overall participation rate remains a percentage point below the CBO’s 2011 projections so it is now short two million workers; many more workers have left the labor force compared with CBO projections from just five years back (see Figure 2). Looking within eight demographic groups (by gender and age), if participation was at each group’s 1994 rate, there would be six million additional workers and the participation rate would be 66.0 percent instead of 63.7 percent. Even this underestimates the labor-force slack. If the labor-force participation rate for each of these groups was as high as the highest rate for that group between 1970 and 2016, then there would be nearly seventeen million additional workers and the labor-force participation rate would be 70 percent.
Adult labor-force participation rate, age 16+, 1960 to 2016.
Although a participation rate of 70 percent would be consistent with past American experience, it would only place the United States to the middle rank among affluent countries. Other Organisation for Economic Co-operation and Development (OECD) members achieve higher labor-force participation rates through aggressive demand policies as well as labor market policies, ranging from effective public employment services, job training programs, and policies to support working mothers (Kleven 2014). Only politics prevents the United States from adopting similar such policies, and achieving greater labor-force participation and capacity output, and the political will was exactly what I explored in my paper on Sanders.
The low rate of labor-force participation in the United States compared with other affluent OECD members suggests that there is considerable space for higher growth rates through increasing employment (see Figure 3). Even at the highest labor-force participation rates seen since 1970, the rates for age and gender groups in the United States are lower than in most of these seven other countries. Compared with the country with the highest labor-force participation rate for each of these three age groups, the United States is today short by 12 percentage points for the young, 10 points for adults, and 15 points for the elderly (see Figure 4).
Labor-force participations by age, United States and seven other OECD members, 2015.
US labor-force participation rate, shortfall compared with highest rate of eight affluent OECD member countries, by age group, 2015.
2.2. Population Size and Immigration
The United States could also grow faster because faster growth in employment would lead to faster population growth through increased immigration. Throughout American history, immigration has been strongly procyclical and thus has served as a source of increasing labor supply during economic expansions (Jerome 1926). Although population aging is determined by past births, mortality, and time, this static view of labor-force growth as coming from natural demographic conditions misses much of the growth in the American labor force because the United States benefits from the regular inflow of millions of additional workers through migration, both documented and undocumented. In 2016, immigrants comprise over 13 percent of the population, and because the labor-force participation rate is higher for immigrants, a larger share (nearly 17 percent) of the workforce. For the decade before the Lesser Depression, immigration (documented and other) accounted for nearly half of the increase in the labor force. The CBO recognizes the cyclical nature of immigration especially that of the undocumented, but, again, discounts the impact of a strong economic recovery program.
3. Direct Effects of Demand on Productivity
In the shadows of exogenous growth theory, there have remained economists who have studied the ways that activist policy can have long-term effects by raising productivity (Dixon and Thirlwall 1975; Millemaci and Ofria 2012; Verdoorn 2002). Writing (in Italian) shortly after the Great Depression and World War II, the Dutch economist Petrus Verdoorn estimated that productivity growth accounts for nearly half of increased output during booms. Verdoorn suggested two sources for this productivity effect: the shift of labor from less productive to more productive industries, and technological progress associated with larger scale of production. Channeling Adam Smith, he said that “one would have expected a priori to find a correlation between labour productivity and output, given that the division of labour only comes about through increases in the volume of production; therefore the expansion of production creates the possibility of further rationalization” (quoted in Boulier 1984: 259).
Verdoorn’s hypothesis was applied by Nicholas Kaldor to explain slow productivity growth in postwar Britain (Kaldor 1966). Kaldor and others later added to labor migration and economies to scale the productivity gains associated with learning by doing (Thirlwall 2003). Recent empirical studies have confirmed that labor productivity is procyclical with an elasticity as high as 0.5 to 0.6 (Castiglione 2011; Jeon and Vernengo 2008). This does not necessarily establish that the long-term growth rate of productivity increases with increasing demand and capacity utilization. It may be that productivity falls in the business downturn because of labor hoarding and the acceleration comes when those hoarded workers are again fully employed (Ball, Leigh, and Loungani 2013; evidence that this is not the case is presented by Martin, Munyan, and Wilson [2015]).
There are several routes through which we may have lasting productivity gains from a fully employed, high-demand economy. In addition to procyclical investment, including research and innovation, productivity will grow from lasting gains from experience and learning. Indeed, this is the idea behind the work of Adam Smith, the first exponent of Verdoorn’s law. Adam Smith argues that “[t]he greatest improvement in the productive powers of labour. . . have been the effects of the division of labour” and that these effects are limited “by the extent of the market” (Smith 1776: 8, 18). Beyond the gains from low-cost transportation and extended trade, Smith is arguing for market-widening effective demand policies. Beyond economies to scale there are economies of agglomeration as well as learning effects associated with greater experience. “A common smith, who, though accustomed to handle the hammer, has never been used to make nails,” Smith notes, “could scarcely make above two or three hundred nails in a day, and those too very bad ones.” Such is the effect of practice, however, that a “smith who has been accustomed to make nails” could make “upwards of two thousand three hundred nails in a day” (Smith 1776: 11). Add to this the effect on technological progress of the division of labor:
[T]he invention of all those machines by which labour is so much facilitated and abridged, seems to have been originally owing to the division of labour. Men are much more likely to discover easier and readier methods of attaining any object, when the whole attention of their minds is directed towards that single object, than when it is dissipated among a great variety of things. (Smith 1776: 12)
I dwell on Smith to emphasize that the idea that demand-led growth will have long-term productivity benefits is as old as the Economics profession. When called learning by doing, demand-led innovation and growth coming from the extension and practice of the division of labor has been shown to lead to major increases in productivity. Not only does practice make violinists ready for Carnegie Hall, it also makes workers better at their tasks and facilitates the discovery and implementation of the abundance of little technological innovations that promote long-term growth. “Learning,” Kenneth Arrow argued half a century ago, “is the product of experience. Learning can only take place through the attempt to solve a problem and therefore only takes place during activity” (Arrow 1962: 155). Learning to be more productive only happens when people are working, and when they are working, learning happens.
There was a time when economists understood this and appreciated the long-term effects on capacity output of demand-led growth. The lasting benefits of demand-led learning have been shown in a variety of settings from Swedish steel manufacture to American textiles (David 1975). The benefits of practice in a high-pressure economy were starkly revealed during World War II when American industry accomplished wonders of productivity, increasing the production of war material more than tenfold even while the labor force shrank and while maintaining higher levels of consumer spending.
These wartime feats of productivity were achieved with technological innovations achieved through rapid learning by doing. Over 2,500 Liberty Ships were built assembly-line style from components made in different locations and then assembled at the dry dock. This novel technology allowed greater specialization than had ever been seen before, and, as Adam Smith might have predicted, greater learning and increased dexterity. The first Liberty Ship required 239 days, the average fell to 44 days and one, the Robert E. Peary, was built in less than five days. Productivity increased by as much as 24 percent for each doubling of output, allowing an increase in productivity of 40 percent per annum (Rapping 1965). Technological improvements developed in the Liberty Ship program carried over elsewhere, including a process to roll cold steel, and the use of welds instead of rivets (Baime 2014; Kennedy 2013).
Practice and learning raised productivity in other parts of the war economy as well. Armen Alchian found learning gains in airframe construction very similar to those reported by Rapping for Liberty Ships (Alchian 1963). Perhaps the most dramatic story involved the world’s largest aircraft manufacturer, and the largest building. It took over a year’s practice, as well as significant investments in R&D, worker training, and physical plant, to make the Ford Motor Company’s Willow Run factory work. But by 1943, B24 bombers were rolling off the plant’s assembly line at a pace of one an hour (Baime 2014).
4. Room for Growth?
How fast could the US economy grow? Compared with the projections by the CBO, I estimate that over the next decade, enactment of the Sanders program could raise the annual growth rate from 1.7 percent a year to 4.5 percent (see Figure 5). Although greater than recent growth rates, this would be slower than the average annual growth rate in economic expansions in the twentieth century.
Annual GDP growth, Friedman estimates of impact of the Sanders program versus CBO projections and past experience.
Some of the higher growth comes from reversing the output decline in the Lesser Depression (Robert Hall estimates that the Lesser Depression lowered output in 2013 by 10 % Hall 2014). Some comes from bringing labor-force participation back to past levels, and up to world levels. The rest comes from boosting productivity growth by 1 percent a year with higher investment and greater learning by doing, reversing the harmful effects of decades of slow growth.
1
Narayana Kocherlakota, President of the Minneapolis Federal Reserve Bank, argued that this is exactly what happened with the New Deal and during World War II. Total factor productivity in 1933 was 15 percent below trend but it grew so rapidly during the New Deal expansion that it had returned to its pre-Depression trend by 1936 (Kocherlakota 2014). Kocherlakota concluded:
[T]here are good reasons to believe that policies that would generate super-normal demand growth in the next four years would also lead to super-normal TFP growth. Taking into account my caveat above, I would (very cautiously) offer the estimate of an additional 1 percent per year for this latter effect. Overall: I see it as possible and beneficial to adopt policies that would lead to
Footnotes
Acknowledgements
I am grateful to Michael Ash, Daniele Girardi, Debra Jacobson, Ted Levy, Antonella Palumbo, Mark Paul, Simon Sturn, Mark Stelzner, and Attilio Trezzini, and to seminar participants at the New School University and the University of Rome Tri for comments and corrections. I am especially grateful to Christina Romer, David Romer, and Justin Wolfers for their trenchant and honest criticisms. Mistakes remain my own.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
1
My view also differs from DeLong’s here, in the assessment of the performance of the US economy before the Lesser Depression. DeLong focuses on the hysteresis effects of the Lesser Depression and assesses my claims for capacity growth in terms of recovery from that shock. I would argue, instead, that the Sanders program was intended not only to promote recovery but also to reverse the debilitating effects of slow economic growth since the 1970s, with a long-term slowdown in investment, labor force, and productivity growth.
Author Biography
Professor of Economics at the University of Massachusetts,